In modern packet-based cellular mobile communication systems, Dynamic Channel Assignment (DCA) schemes are popular, since they are an efficient tool to increase the (air interface) system throughput. DCA schemes utilize the short term fluctuations (fast fading) of the channel quality of the links between base stations (BS) and mobile stations (MS). In such a system a so-called scheduler (usually part of the base station) tries to assign system resources preferably to mobile stations in favorable channel conditions.
In time domain DCA works on a frame-by-frame basis, where a frame duration is typically in the (sub-)millisecond region. Furthermore—depending on the multiple access scheme—the air interface resources are divided in e.g. code and/or frequency domain.
The following description concentrates on downlink scenarios (BS transmits to MS), however without loss of generality, DCA can also be applied to the uplink (MS transmits to BS). In any case, the scheduler performing the DCA needs to have detailed channel knowledge of the BS-MS links, which is gathered by channel estimation. If the scheduler is located in the network and the measurement is performed in the MS, the channel information is signaled from MS to BS. It is important, that the channel quality is measured on a instantaneous basis in order to reflect the instantaneous received signal power and the instantaneous interference.
In Frequency Division Multiple Access (FDMA) systems, DCA is performed in time-frequency domain, since physical layer channels are defined in frequency domain. Typically, the channel quality varies significantly in frequency domain (frequency selective fading). Hence, depending on the conditions of the channels over all available frequencies and all active mobile stations, the scheduler can assign the channels dynamically at each scheduling instant to specific BS-MS links.
In an OFDMA (Orthogonal Frequency Division Multiple Access) system, the frequency resource is partitioned into narrowband subcarriers, which typically experience flat fading. Here, generally the scheduler dynamically assigns subcarrier blocks (containing M adjacent or separated subcarriers) to a specific MS in order to utilize favorable channel conditions on a link. Example of such a system is known from Rohling et al., “Performance of an OFDM-TDMA mobile communication system”, IEEE Proceedings on the Conference on Vehicular Technology (VTC 1996), Atlanta, 1996.
In case of a CDMA (Code Division Multiple Access) the system resources are defined in code domain and, therefore, the scheduler dynamically assigns codes to specific BS-MS links. Note, that in contrast to FDMA, for a given link the channel quality is similar for all resources/codes (fading is not code selective) and, hence, in code domain the DCA is performed with respect to the number of codes to assign to a specific MS and not which codes to assign. The DCA is focused on the time domain scheduling utilizing the fast fading characteristics. HSDPA (High Speed Downlink Packet Access) within the 3GPP (3rd Generation Partnership Project) standard is such a CDMA system employing DCA.
A MC-CDMA (Multi-Carrier CDMA) system can be considered as a combination of CDMA and (O)FDMA. Hence, DCA can be performed as well in code as in frequency domain.
Generally, the DCA throughput efficiency increases with the number of active mobile stations in a cell, since this increases the number of links in good channel conditions and, therefore, increases the probability that a channel in favorable conditions is scheduled (multi-user diversity).
Typically, DCA is combined with link adaptation techniques such as Adaptive Modulation and Coding (AMC) and hybrid Automatic Repeat reQuest (ARQ).
Furthermore, DCA can be combined with power control schemes, where the power assigned to a specific channel (in code, frequency domain) is controlled in order to compensate the channel power variations and/or to support the AMC operation.
Systems without Power Control
As described in the previous section, for efficient DCA operation the scheduler in the BS when assuming a non-power controlled system needs detailed knowledge on the instantaneous quality of all channels over all available subcarrier blocks and all involved BS-MS links.
Considering a DCA OFDMA multi-cell scenario and a frequency re-use factor of 1, the system is typically interference limited. I.e. the channel quality per subcarrier block is primarily defined by the signal (S) to interference (I) ratio (SIR), where the interference is dominated by the intercell-interference (co-channel interference) caused by the transmissions on the respective channel (subcarrier block) in adjacent cells (C denotes the set of adjacent cells):
                              ChannelQuality          ≈                      S            ⁢                                                  ⁢            I            ⁢                                                  ⁢            R                          =                              S            I                    ≈                      S                                          ∑                C                            ⁢                              I                c                                                                        (        1        )            
In case of an OFDMA system with DCA and frequency selective fading, the instantaneous SIR(t) for a given link to a mobile station m varies over the subcarrier blocks b, since both the signal and the interference experience fading:
                              S          ⁢                                          ⁢          I          ⁢                                          ⁢                                    R              b              m                        ⁡                          (              t              )                                      =                                                            S                b                m                            ⁡                              (                t                )                                                                    I                b                m                            ⁡                              (                t                )                                              ≈                                                    S                b                m                            ⁡                              (                t                )                                                                    ∑                C                            ⁢                                                (                                                            I                      b                      m                                        ⁡                                          (                      t                      )                                                        )                                c                                                                        (        2        )            
As mentioned earlier, the performance of a system employing DCA and AMC greatly depends on the accuracy of the SIR estimation. Therefore, according to equation (2) the following problems occur:
All values in equation (2) experience fast fading and will change between the point in time of the measurement and the point in time of the actual transmission (after performing DCA and AMC selection). This delay causes inaccurate DCA and AMC operation. The delay even increases, if the measurement is performed at the MS and needs to be fed back by signaling to the BS.
The number of interferers in the denominator depends on the actual usage (allocation) of the subcarrier block in the adjacent cells. I.e. depending on the actual load in the adjacent cells some subcarrier blocks might not be used. Generally, at the point in time of the measurement, the usage of subcarrier block at the point in time of the transmission is unknown in adjacent cells due to the following reasons:
The channel quality measurement is performed based on an outdated interference caused by the subcarrier block allocation (scheduling) in the adjacent cells (measurement for the n-th frame is performed at the (n-k)-th frame, where the subcarrier allocation is most likely different).
Further, there exists the so-called chicken-and-egg allocation problem: In cell A, the subcarrier block allocation and AMC can only be performed after the SIR measurement/calculation in cell A has been performed, which requires knowledge of the subcarrier block allocation in cell B (adjacent cells). However, before the subcarrier block allocation in cell B can be performed the SIR measurement/calculation in cell B needs to be performed, which requires the knowledge of the subcarrier block allocation in cell A.
In case the chicken-and-egg problem may be avoided/solved by e.g. an iterative process, signaling of e.g. the allocation status between base stations would be required. However, since the scheduling frames are in the millisecond region, the signaling would introduce additional significant delay.
Additionally, without any power control, the average SIR (neglecting fast fading influences) for a BS-MS link strongly depends on the geometry (e.g. distance to BS) of the MS causing the following effects:
With increasing distance between BS and MS, the SIR for the respective links decreases, since the average received signal power decreases and the average received interference power increases. This translates in a significantly lower achievable data rate per subcarrier-block for links to mobile stations in low geometry.
The difference in average SIR can be on the order of tens of dB, which requires a large dynamic range for the AMC scheme definition. This leads to an increased amount of signaling, since the required number of combinations of modulation schemes and code rates increases when keeping the AMC granularity with respect to smaller dynamic ranges.
Compared to power controlled systems, in non-power controlled systems it is more likely that multilevel modulation schemes (e.g. 8-PSK, 16-QAM, 64-QAM, etc) are chosen for links to mobile stations in high geometry. Although, this increases the available throughput for those mobile stations, it can decrease the overall system throughput compared to a system, where the available power is distributed such that only non-multilevel modulation schemes (e.g. QPSK) are used. This is caused by the reduced power efficiency of multilevel modulation schemes.
Further, in contrast to power controlled systems, in non-power controlled systems it is more likely that mobile stations in low geometry cannot receive any data with single transmission attempts, but would need several retransmissions. Therefore, the average number of transmissions (ARQ retransmissions) increases, which in turn increases the transmission delay and feedback signaling, as well as decreasing the bandwidth efficiency.
Data transmission to mobile stations in high geometry is burstier in the time domain, since on average higher modulation and coding schemes can be selected. This results in a burstier subcarrier block allocation. This will make the SIR estimation according to equation (2) more difficult, since the subcarrier block allocation changes more often.
Properties of Power Controlled Systems
DCA and AMC can also be combined with Power Control (PC) schemes. Employing PC the system tries to compensate fluctuations of the received signal power due to the signal path loss, shadowing effects (slow fading) and or fast fading effects. Generally, PC schemes can be classified into two categories: Fast PC and slow PC.
In contrast to systems without PC, for slow PC systems the average SIR does not depend on the geometry of the mobile stations, assuming only slow fading effects and unlimited minimum and maximum transmit power. Hence, the achievable data rates per subcarrier block do not depend on the MS position. Note however, the slow PC can only operate within certain limits (dynamic range of the control commands), i.e. the power compensation might not be sufficient or fast enough for any link.
Fast power control is usually performed jointly with the AMC in order to adapt the transmission rate to short term fluctuations and in order to optimize the overall power usage.
With slow/fast PC the instantaneous SIR estimation/measurement/calculation problem as outlined in the previous sections above, is more severe compared to the non-PC case. That is, the unknown number of interference components of the sum in the denominator equation (2) do not only experience fast fading, but significantly vary in amplitude due to the PC in adjacent cells. I.e. the intercell-interference on a given subcarrier block from a given adjacent cell can vary from frame to frame in tens of dB depending on which MS is scheduled on the respective subcarrier block, since the transmitted power might vary significantly depending primarily on the MS location. This is especially critical, if the interference is dominated by few interferers, since there is no interference averaging effect.